atomic structure of the apoptosome: mechanism of cytochrome c...

Atomic structure of the apoptosome:mechanism of cytochrome c- anddATP-mediated activation of Apaf-1Mengying Zhou,1,3 Yini Li,1,3 Qi Hu,1 Xiao-chen Bai,2 Weiyun Huang,1 Chuangye Yan,1

Sjors H.W. Scheres,2 and Yigong Shi1

1Ministry of Education Protein Science Laboratory, Center for Structural Biology, Tsinghua-Peking Center for Life Sciences,School of Life Sciences, School of Medicine, Tsinghua University, Beijing 100084, China; 2MRC Laboratory of Molecular Biology,Cambridge Biomedical Campus, Cambridge CB2 0QH, United Kingdom

The apoptotic protease-activating factor 1 (Apaf-1) controls the onset of many known forms of intrinsic apoptosis inmammals. Apaf-1 exists in normal cells as an autoinhibited monomer. Upon binding to cytochrome c and dATP,Apaf-1 oligomerizes into a heptameric complex known as the apoptosome, which recruits and activates cell-killingcaspases. Here we present an atomic structure of an intact mammalian apoptosome at 3.8 Å resolution, determinedby single-particle, cryo-electron microscopy (cryo-EM). Structural analysis, together with structure-guided bio-chemical characterization, uncovered how cytochrome c releases the autoinhibition of Apaf-1 through specific in-teractions with the WD40 repeats. Structural comparison with autoinhibited Apaf-1 revealed how dATP bindingtriggers a set of conformational changes that results in the formation of the apoptosome. Together, these resultsconstitute the molecular mechanism of cytochrome c- and dATP-mediated activation of Apaf-1.

[Keywords: apoptosome; cryo-EM structure; apoptosis; Apaf-1; caspase activation; caspase-9]

Supplemental material is available for this article.

Received September 20, 2015; revised version accepted October 13, 2015.

Programmed cell death (apoptosis) is essential for the de-velopment and tissue homeostasis of all multicellularorganisms (Horvitz 2003; Danial and Korsmeyer 2004).Apoptosis is executed by sequential activation of initiatorand effector caspases (Shi 2002; Yan and Shi 2005). Themajor function of effector caspases, exemplified by themammalian caspase-3, is to kill a cell by inflicting numer-ous cleavages on life-sustaining proteins. The primary roleof an initiator caspase, on the other hand, is to cleave andhence activate specific effector caspases. In mammaliancells, the initiator caspase-9 is responsible for the activa-tion of caspase-3. The autocatalytic activation of an initi-ator caspase depends critically on a specific, multiproteincomplex. For caspase-9, this multiprotein complex isknown as the apoptosome, which comprises seven copiesof a heterodimer between apoptotic protease-activatingfactor 1 (Apaf-1) and cytochrome c (CytC) (Chai and Shi2014). Because caspase-9 is essential for most knownforms of intrinsic apoptosis, elucidation of its activationmechanism has been a central task toward mechanistic

understanding of programmed cell death. The first majorstep toward this goal is to elucidate the mechanism ofapoptosome assembly.In normal mammalian cells, Apaf-1 exists as an ADP-

bound, autoinhibited monomer. In response to variousforms of intrinsic cell death stimuli, CytC is releasedfrom mitochondria into the cytoplasm (Liu et al. 1996),where CytC binds the monomeric Apaf-1 and primes itfor oligomerization (Li et al. 1997; Zou et al. 1997). Thereplacement of ADP by dATP or ATP results in markedconformational changes, allowing Apaf-1 to form an acti-vated, heptameric apoptosome (Acehan et al. 2002; Kimet al. 2005; Bao et al. 2007). Only the activated apopto-some is capable of facilitating the autocatalytic activationof caspase-9 (Rodriguez and Lazebnik 1999; Saleh et al.1999; Zou et al. 1999). How does CytC interact withApaf-1? How can these interactions facilitate nucleotideexchange and oligomerization of Apaf-1? What is themechanism of apoptosome-mediated activation of cas-pase-9? Despite rigorous investigations, these questionshave remained largely enigmatic.The cryo-electron microscopy (cryo-EM) structures of

the Apaf-1 apoptosome have been elucidated at a resolu-tion range between 9.5 and 21 Å (Acehan et al. 2002; Yu

3These authors contributed equally to this work.Corresponding authors: [email protected], [email protected] published online ahead of print. Article and publication dateare online at http://www.genesdev.org/cgi/doi/10.1101/gad.272278.115.Freely available online through the Genes & Development Open Accessoption.

© 2015 Zhou et al. This article, published in Genes & Development, isavailable under a Creative Commons License (Attribution 4.0 Internation-al), as described at http://creativecommons.org/licenses/by/4.0/.

GENES & DEVELOPMENT 29:2349–2361 Published by Cold Spring Harbor Laboratory Press; ISSN 0890-9369/15; www.genesdev.org 2349

Cold Spring Harbor Laboratory Press on October 13, 2020 - Published by genesdev.cshlp.orgDownloaded from

mailto:[email protected]








http://www.genesdev.org/cgi/doi/10.1101/gad.272278.115

http://www.genesdev.org/cgi/doi/10.1101/gad.272278.115

http://genesdev.cshlp.org/site/misc/terms.xhtml

http://creativecommons.org/licenses/by/4.0/


http://genesdev.cshlp.org/site/misc/terms.xhtml

http://genesdev.cshlp.org/

http://www.cshlpress.com

et al. 2005; Yuan et al. 2010, 2011). These structures al-lowed placement of individual domains but failed to re-veal specific interactions that govern the function of theapoptosome. X-ray structures of the monomeric, ADP-bound Apaf-1 provided an atomic view of the underpin-nings of Apaf-1 autoinhibition (Riedl et al. 2005; Reuboldet al. 2011). Together, these structural observations al-lowed the proposition of a speculative model that delin-eates the assembly of the apoptosome (Yuan et al. 2013).This model suffers from relatively low resolutions of priorEM structures and lack of supporting biochemical data.

In this study, we report the three-dimensional (3D)structure of an intact Apaf-1 apoptosome at a near-atomicresolution of 3.8 Å, determined by single-particle, cryo-EM analysis. We present results of structure-guided bio-chemical analyses. These experimental data give rise toa mechanistic pathway of Apaf-1 activation and apopto-some assembly.

Results

Overall structure of the Apaf-1 apoptosome

The full-length human Apaf-1 was expressed in baculovi-rus-infected insect cells and biochemically purified to ho-mogeneity. Assembly of an intact apoptosome wascompleted through incubation of the purified Apaf-1 pro-tein with an excess amount of horse CytC and 1 mMdATP. The assembled apoptosome exhibited excellentsolution behavior on gel filtration (Supplemental Fig.S1A) and markedly stimulated the proteolytic activity ofcaspase-9. We imaged the apoptosome sample undercryo-conditions on an FEI Titan Krios microscope operat-ing at 300 kV and collected 912 micrographs (Supplemen-tal Fig. S1B). We chose a total of 202,932 individualparticles for reference-free two-dimensional (2D) classifi-cation (Supplemental Fig. S1C). After 3D classification, asubset of 134,919 particles was used for image construc-tion, giving a final overall resolution of 3.8 Å on the basisof the gold Fourier shell correlation (FSC) standard (Sup-plemental Figs. S1D, S2).

The overall architecture of the Apaf-1 apoptosome issimilar to that reported previously (Yuan et al. 2013), com-prising a central hub and seven spokes (Fig. 1A). The localresolution for the vast majority of the central hub rangesbetween 3.0 and 3.5 Å (Fig. 1A), which represents a quali-tative improvement over previously reported resolutionsand, for the first time, allows assignment of specific inter-actions involving amino acid side chains in the activatedapoptosome (Supplemental Fig. S2). The local resolutionat the spokes is considerably lower (Fig. 1A); local mask-ing strategy markedly improved the resolution in this re-gion to ∼5 Å, which allows unambiguous determinationof the interface between Apaf-1 and CytC (Fig. 1B).

Apaf-1 contains a caspase recruitment domain (CARD)at the N terminus, a nucleotide-binding domain (NBD), ahelical domain (HD1), a winged helix domain (WHD), asecond helical domain (HD2), and 15 WD40 repeats attheC-terminal half. The central hub consists of seven cop-ies each of the NBD, HD1, and WHD (Fig. 1C). The 15

WD40 repeats in each spoke constitute two β propellersnamedWD1 andWD2,which are connected to the centralhub through the HD2 domain. The seven CARDs exhibitno EM density, likely reflecting their dynamic locations.CytC is sandwiched by the front faces of the two β pro-pellers. The CytC-exposed side of the apoptosome is here-after referred to as the top face (Fig. 1C), which ischaracterized by a ring of positively charged residues atthe surface of the central hub (Supplemental Figs. S3,S4). In contrast, the bottom face of the apoptosome is en-riched by negatively charged amino acids (SupplementalFigs. S3, S4).

Compared with the reported structure of the apopto-some at 9.5 Å resolution (Yuan et al. 2010), our structureat 3.8 Å reveals a number of novel findings. First, the ob-served docking interface between the two β propellersand CytC in our structure is different from that proposedpreviously (Yuan et al. 2010, 2013). Compared with thepublishedmodel (Yuan et al. 2010, 2013), CytC undergoesa rotation of ∼90° (Supplemental Fig. S5). This structuralrevelation has important ramifications for understanding

Figure 1. Overall structure of the Apaf-1 apoptosome. (A) Anoverall view of the EM density for the Apaf-1 apoptosome. Theresolution is color-coded for different regions of the apoptosome.The surface view of the apoptosome is shown here. The resolu-tion goes to 3.0–3.5 Å in the central hub of the apoptosome. (B)Two close-up views of the EM density surrounding CytC. Thetwo β propellers WD1 and WD2 consist of WD40 repeats 1–7and 8–15, respectively. (C ) Overall structure of theApaf-1 apopto-some. Two views are shown. The top face refers to the CytC-ex-posed side of the apoptosome disk. CytC is colored yellow, andthe domains within each Apaf-1 protomer are color-coded. Aand Bwere prepared using Chimera (Pettersen et al. 2004). Exceptfor Figures 5 and 6, all other figures were prepared using PyMol(http://www.pymol.org).

Zhou et al.

2350 GENES & DEVELOPMENT


http://www.pymol.org







how CytC releases the autoinhibition of the monomericApaf-1. Second, nucleotide binding is observed for thefirst time in the activated Apaf-1 apoptosome, and dATPis coordinated by several key residues through a numberof specific hydrogen bonds (H bonds; discussed later).Third and most importantly, the markedly improvedresolution allows visualization of atomic features and spe-cific interactions, which govern Apaf-1 activation andoligomerization.

Intramolecular domain stacking

Our cryo-EM structure of the apoptosome represents thefirst experimentally determined atomic model of the acti-vated Apaf-1, which exhibits a dimension of ∼145 Åin height, 80 Å in width, and 55 Å in thickness (Fig. 2A).The overall appearance of Apaf-1 resembles a seahorse,with NBD/HD1 and WD1/WD2/CytC corresponding tothe head and the tail, respectively (Fig. 2A). The headand the tail are spatially separated by a gap of 12–14 Åand are connected by the WHD and HD2 domains. Thesurface of Apaf-1 is enriched by charged amino acids(Fig. 2A), which engender specific intermolecular H bondsand salt bridges between neighboring Apaf-1 protomers inthe apoptosome (discussed later). The extended structureof Apaf-1 is held together through extensive interdomaininteractions, mainly involving three interfaces: betweenWHD and NBD/HD1 (Fig. 2B), between WHD and HD2(Fig. 2C), and between HD2 and WD1/WD2 (Fig. 2D).The WHD stacks against the NBD/HD1 module

through a combination of H bonds and van der Waalscontacts (Fig. 2B). The periphery contains at least eightpotential H bonds, exemplified by four pairs of aspartateand lysine residues. The charged side chains of Asp365,Lys421, Asp439, and Asp443 from WHD interact with

the side chains of Lys351, Asp266, Lys318, and Lys348, re-spectively. At the center of this interface, the hydrophobicside chains of Leu364,Met368, and Leu440 fromWHDarenestled in a shallow hydrophobic pocket formed by sidechains from Met155/Ala156 from NBD and Leu322/Val323/Leu326/Phe350 from HD1 (Fig. 2B). In contrastto the interface between WHD and NBD/HD1, WHD in-teracts with HD2 through predominantly van der Waalscontacts (Fig. 2C). Notably, Ile388, Val393, Val399, andLeu403 from WHD closely stack against a hydrophobicsurface patch formed by His457, Ile460, His490, and fouraromatic residues: Tyr482, Trp483, Phe486, and Tyr489.The interface betweenHD2 and theWD1/WD2module

involves a large buried surface area of ∼3362 Å2, of whichapproximately two-thirds are contributed by the WD2propeller (Fig. 2D, top panel). In particular, four hydropho-bic residues—Leu593, Leu595, Trp597, and Ile603—fromHD2 follow a hydrophobic corridor on the surface ofWD2 that is formed by 11 uncharged residues (Fig. 2D,bottom panel). This high density of van der Waals con-tacts between HD2 and WD2 is reminiscent of that be-tween WHD and HD2 but contrasts the relatively sparseinteractions betweenHD2 andWD1. This analysis is fullyconsistent with the notion that the three domains WHD,HD2, and WD2 form a rigid rod (Reubold et al. 2011). TheWHD/HD2/WD2 rod remains unchanged in response toApaf-1 activation (discussed later).

Interactions among adjacent Apaf-1 protomersin the apoptosome

In the apoptosome, two adjacent Apaf-1 protomers inter-act with each other exclusively through their N-terminalregions (Fig. 3A). Specifically, the NBDs andWHDs of oneprotomer stack closely against the NBD and HD1,

Figure 2. Structure of an activated Apaf-1 proto-mer in the apoptosome. (A) Structure of an activat-ed Apaf-1 protomer in the apoptosome is shown incartoon representation (left panel) and electrostaticsurface potential (right panel). Three boxed inter-domain interfaces are detailed in B–D. (B) Aclose-up view of the interface between WHD andthe NBD–HD1 module. Potential hydrogen bondsare represented by red dashed lines. (C ) A close-up view of the interface between WHD and HD2.This interface contains a large number of van derWaals contacts. (D) A close-up view of the interfacebetween HD2 and the two β propellers. The inten-sity of interaction between HD2 and WD2 is con-siderably stronger than that between HD2 andWD1. The bottom panel zooms in on the van derWaals interactions mediated by three hydrophobicresidues from HD2: Leu595, Trp597, and Ile603.

Atomic structure of the Apaf-1 apoptosome

GENES & DEVELOPMENT 2351




respectively, of an adjacent protomer (Fig. 3B). This modeof interaction, characterized by a prominent role of theWHD, is uniquely shared among the NOD family of pro-teins (Qi et al. 2010; Pang et al. 2015) and differs fromthat for most other AAA+ motors (Diemand and Lupas2006; Erzberger and Berger 2006). Our cryo-EM structureat 3.8 Å resolution allows identification of interprotomerinteractions that govern the assembly of the activatedapoptosome.

The stacking interactions between two adjacent Apaf-1protomers involve a buried surface area of ∼2805 Å2, ofwhich three-quarters are contributed by the interface be-tween two neighboring NBDs. This interface is promi-nently configured by 13 potential H bonds (Fig. 3C). Atthe center of the interface, the carboxylate side chainof Asp225 accepts an intermolecular H bond from theguanidinium group of Arg205, whereas the guanidiniumgroup of Arg233 donates two intermolecular H bondsto Glu116 and Gln208. At one side of the periphery,Glu146 and Gln257 each accept an intermolecular Hbond from Arg111 and Gln121, respectively. At the otherside, a network of eight H bonds is placed close to eachother and may play a key role in stabilizing the interface(Fig. 3C).

The interface between the WHD of one Apaf-1 proto-mer and HD1 of an adjacent protomer is also enrichedby charged amino acids (Fig. 3D) but appears to play a sup-porting role for apoptosome assembly, as judged by amuch smaller buried surface area of ∼763 Å2 comparedwith that between two adjacent NBDs. Consistent withthis analysis, a truncated Apaf-1 without the WHD stillretained the ability to form an oligomer (Hu et al. 1998;Srinivasula et al. 1998). This interface features five inter-

molecular H bonds (exemplified by a bifurcated contactfrom Arg337 of HD1) and van der Waals interactions be-tween Phe433 and Phe334 (Fig. 3D).

The intermolecular interactions among neighboringApaf-1 protomers comprise mostly charge-stabilized Hbonds. This structural observation sharply contrasts theinterdomain interactionswithin a singleApaf-1 protomer,which are dominated by van der Waals contacts (Fig. 2).Compared with hydrophobic interactions, H bonds andsalt bridges usually exhibit faster kinetics and thus aremore amenable to biological regulation—in this case,apoptosome assembly and disassembly for the activationof caspase-9. Therefore, the atomic interactions bothwithin an Apaf-1 protomer and between adjacent proto-mers appear to have evolved to facilitate the underlyingfunctions.

Recognition of CytC by Apaf-1

In contrast to the central hub, the periphery of the Apaf-1apoptosome, comprising theWD40 repeats and CytC, dis-plays relatively low resolution. This feature may reflectthe inherent property of the apoptosome: rigid conforma-tion at the central hub and relatively mobile nature in thespokes. The strategy of applying local masks helped im-prove the EM density and allowed unambiguous determi-nation of CytC binding. CytC is sandwiched between thetwo front faces of WD1 and WD2, with excellent shapeand charge complementarity (Fig. 4A). Although the sec-ondary structural elements in CytC and WD1/WD2 andtheir relative orientation are well defined (Fig. 1B), theimproved EM density is inadequate for assignment of spe-cific side chains. Nonetheless, we manually modeled

Figure 3. Interactions between two adjacentApaf-1 protomers in the apoptosome. (A) An overall viewof two adjacent Apaf-1 protomers in the apopto-some. The top face is shown. The domains are col-or-coded. (B) A focused view of the domains thatmediate intermolecular interactions. An Apaf-1protomer uses its NBD and WHD to contact theNBD and HD1, respectively, of an adjacent Apaf-1protomer. The interactions are detailed in C andD. (C ) A close-up view of the interface betweentwo neighboring NBDs. This interface contains alarge number of H-bonds (red dashed lines). An in-set is shown to highlight a network of closelyspacedH bonds. (D) A close-up viewof the interfacebetween the WHD of one Apaf-1 protomer andHD1 of an adjacent Apaf-1 protomer.

Zhou et al.





putative interactions for specific residues between CytCand the WD1–WD2 module in Apaf-1.The interface betweenCytC andWD2ofApaf-1, involv-

ing ∼1242 Å2 buried surface area, appears to contain morespecific interactions than that between CytC and WD1,which entails∼712Å2 buried surface area (Fig. 4B). The in-terface between CytC and WD2 of Apaf-1 contains threeputative intermolecular H bonds and a number of vanderWaals contacts. In particular, Gly56 of CytC is locatedin close proximity to Thr1087 from WD2, with a H bondbetween the side chain of Thr1087 and the carbonyl oxy-gen of Gly56 (Fig. 4C). The side chains of Lys39 and Gln42of CytC may donate two H bonds to the carbonyl oxygenof Phe1063 and the side chain of Glu1045, respectively.In addition, Pro76 of CytC makes direct van der Waalscontacts to the indole ring of Trp1179 from WD2 ofApaf-1 (Fig. 4D). The interface between CytC and WD1of Apaf-1 contains only one putative H bond betweenGln12 of CytC and Glu700 from WD1 (Fig. 4E). Most no-tably, however, Ile81 of CytC is in close contact withTrp884 from WD1 (Fig. 4D). Lys72 of CytC, which hadbeen previously implicated in playing a role in bindingto Apaf-1 (Yu et al. 2001), is likely H-bonded to Asp902from WD1 (Fig. 4D).To corroborate the structural observations, we generat-

ed five representative missense mutations in CytC, indi-vidually purified these mutant proteins to homogeneity,and examined their abilities to associate with Apaf-1.Free wild-type CytC exhibited an elution volume of

∼3.3 mL on gel filtration (Fig. 5A). Upon incubation ofthe full-length human Apaf-1 (residues 1–1248) with anexcess amount of wild-type CytC, the elution volumefor a fraction of wild-type CytC was shifted to ∼2.3 mL,which coincides with that of Apaf-1 (Fig. 5A). This obser-vation suggests that wild-type CytC formed a stable com-plex with Apaf-1. This conclusion was confirmed byisothermal titration calorimetry (ITC), which revealed adissociation constant of ∼0.49 µM± 0.06 µM for theApaf-1–CytC interaction (Fig. 5B). In contrast to wild-type CytC, four CytC mutants—G56K, K72W, P76E,and I81E—abolished stable association with Apaf-1, asjudged by gel filtration analysis (Fig. 5C) as well as ITC(Supplemental Fig. S6). These results nicely corroboratethe structural finding that Gly56, Lys72, Pro76, andIle81 all play an important role in stabilizing the interfacebetween CytC and Apaf-1 (Fig. 4C,D). Lys87 of CytCmakes no direct contact to WD1 or WD2; consequently,the mutation K87W in CytC had little impact on its asso-ciation with Apaf-1 (Fig. 5C; Supplemental Fig. 6).

Impact of CytC mutations on caspase-9 activation

The association of CytCwith Apaf-1 is thought to precedenucleotide exchange and apoptosome formation (Chaiand Shi 2014). Mutations in CytC that weaken its interac-tion with Apaf-1 are predicted to have negative conse-quences on the function of apoptosome. To examinethis scenario, we generated 11 additional missense

Figure 4. Recognition of CytC by the WD1 and WD2domains of Apaf-1. (A) CytC is sandwiched betweenWD1 andWD2with excellent complementarity in shapeand charge. CytC and WD1/WD2 are shown in electro-static surface potential. (B) A schematic diagram of theinteractions between CytC and theWD1/WD2 domains.The three boxed regions are detailed in C–E. (C ) A close-up view of the interface betweenCytC andWD2. This in-terface contains three putative intermolecular H bonds(red dashed lines). Notably, Gly56 contributes a H bondand makes van der Waals contacts to WD2. (D) A close-up view of the interface between CytC and the adjoiningarea of WD2 and WD1. Pro76 and Ile81 make van derWaals contacts to neighboring structural elements fromWD2 and WD1, respectively. Lys72 donates a H bondtoAsp902 ofWD1. (E) A close-up view of the interface be-tween CytC and WD1. Gln12 of CytC likely makes a Hbond to Glu700 of WD1.






mutations in CytC and investigated the ability of all 16CytC mutants to activate caspase-9 using an in vitro pro-teolytic cleavage assay. In this assay, the catalytically in-active caspase-3 (C163A) protein, which is a physiologicalsubstrate of caspase-9, was employed as the substrate. Inthe absence of CytC, caspase-9 had little proteolytic activ-ity toward the substrate (Fig. 6A, lanes 2,4–7). In the pres-ence of Apaf-1 and dATP, wild-type CytC markedlystimulated the protease activity of caspase-9 (Fig. 6A,lane 9). In contrast, five CytC missense mutants (G56K,K72E, P76E, I81E, and K86E), each involving mutationto a charged amino acid, exhibited markedly decreasedlevels of protease activity (Fig. 6A, lanes 13,15,17,21,24).For three of the five targeted residues (Lys72, Pro76,

and Lys86), mutation to Trp was also generated; intrigu-ingly, all threemutants retained a similar level of proteaseactivity compared with wild-type CytC (Fig. 6A, lanes14,16,23). These observations suggest that alteration ofthe surface charge distribution in CytC, especially by in-troducing a charged amino acid into a generally hydropho-bic environment (G56K, P76E, and I81E), is moredetrimental than merely introducing a bulkier residue(Trp). Furthermore, these results indicate that the com-promised Apaf-1 binding by CytC-K72W (Fig. 5C; Supple-mental Fig. S6) had been alleviated in the presence ofdATP. Other CytC mutants also retained a similar levelof protease activity compared with wild-type CytC. To-gether, these experimental findings demonstrate that

Figure 5. Biochemical characterization of theinteractions between Apaf-1 and CytC. (A)Wild-type CytC forms a stable complex withthe full-length Apaf-1. Three gel filtrationchromatograms are shown in the left panel,and the corresponding fractions from gel filtra-tion are shown on SDS-PAGE gels in the rightpanels. To clearly visualize the results, thebaselines for the magenta (CytC alone) andgreen (Apaf-1 +wild-type CytC) chromato-grams were shifted by 25 and 50 mAU, respec-tively. (B) Wild-type CytC binds to the full-length Apaf-1 with an association constant of∼2.04 × 106 M−1 ± 0.25 × 106 M−1, which corre-sponds to a binding affinity of ∼0.49 µM± 0.06µM. Shown here are the raw data from ITC (leftpanel) and curve fitting (right panel). (C ) As-sessment of the interactions between five rep-resentative CytC mutants and Apaf-1 by gelfiltration. To better display the results, thebaselines for the five CytCmutants were shift-ed incrementally by 25 mAU each. Only wild-type CytC and the K87W mutant retained sta-ble association with Apaf-1.

Zhou et al.





the residues in CytC that play a key role at the interfacewith Apaf-1 are indispensable for caspase-9 activation,presumably due to compromised formation of theapoptosome.To further corroborate this conclusion, we switched to

a more sensitive cleavage assay in which the caspase-3substrate was replaced by the fluorogenic peptide Ac-LEHD-AFC (Fig. 6B). This assay allows convenient quan-tification of the protease activity and is more sensitivethan that using the caspase-3 (C163A) protein as the sub-strate (Fig. 6A). Gratifyingly, the results are fully consis-tent with those obtained using caspase-3 (C163A). Thepresence of both wild-type CytC and dATP, but not asingle agent alone, led to a drastically improved cas-pase-9 activity in the presence of Apaf-1. Consistently,

wild-type CytC and Apaf-1 formed a stable apoptosomein the presence of 1 mM dATP (Fig. 6C, top panel). Com-pared with wild-type CytC, the positive control mutantsexemplified by K87E and K87W stimulated caspase-9 ac-tivity to a similar level (Fig. 6B); the CytC mutant K87Walso allowed efficient assembly of the apoptosome (Fig.6C, bottom panel). In sharp contrast to wild-type CytC,the four mutants G56K, K72E, P76E, and I81E nearlyabolished the ability to stimulate caspase-9 activity inthe presence of Apaf-1 (Fig. 6B); accordingly, the CytCmutants G56K, P76E, and I81E abrogated apoptosomeformation (Fig. 6C, bottom panel). The CytC mutantK72W retained the ability to stimulate caspase-9 activity(Fig. 6B) and supported apoptosome formation (Fig. 6C,bottom panel).

Figure 6. Impact of CytC mutations on apoptosome assembly and caspase-9 activation. (A) Impact of CytC mutations on caspase-9 ac-tivation in an in vitro cleavage assay using caspase-3 (C163A) as the substrate. For each CytC mutant, only the residue number and thetarget residue are labeled. For example, “56K” denotes G56K. The only doublemutation (DM) denotes K79F and I81W. (B) Impact of theseCytCmutations on caspase-9 activation in a second in vitro cleavage assay using the fluorogenic peptide Ac-LEHD-AFC as the substrate.(C ) Assessment of the impact of representativeCytCmutations on apoptosome formation by gel filtration. Thewild-type control is shownin the top panel, where the baselines for the green (Apaf-1 alone) and red (CytC alone) chromatograms were shifted by 25 and 50 mAU,respectively. The five representative CytC mutants are displayed in the bottom panel, where the baselines of the chromatograms wereshifted incrementally by 25 mAU each. Notably, the CytC-K72W mutant partially retained the ability to form an apoptosome.






Conformational changes and nucleotide binding

Formation of the apoptosome requires drastic conforma-tional rearrangements in Apaf-1. The relative orientationof NBD versus HD1 remains largely unchanged betweenthe ADP-bound, autoinhibited Apaf-1 and the dATP-bound, activated Apaf-1. Structural alignment aroundthe NBD–HD1 module reveals a pseudo-twofold rotationsymmetry in the relative positioning of the WHD–HD2–WD2 rods (Fig. 7A), which is exemplified by the WHDs(Fig. 7B). Application of this rotation symmetry resultsin near-perfect registry between the two WHD–HD2–WD2 rods (Fig. 7C). In contrast to the WHD–HD2–WD2rods, the two WD1 propellers are positioned quite differ-ently in the two Apaf-1 molecules (Fig. 7D). Comparedwith that in the autoinhibited Apaf-1, the WD1 propellerin the activated Apaf-1 is rotated by ∼60° toward WD2(Fig. 7D), which is apparently caused by CytC binding.Such a drastic conformational change in WD1 would beincompatiblewith the overall conformation of the autoin-hibited Apaf-1 because WD1 would sterically clash withthe NBD (Fig. 7E). This analysis strongly argues thatCytC binding may facilitate the conformational changesin Apaf-1 that destabilize the autoinhibited state.

Thus, the most prominent conformational switch dur-ing Apaf-1 activation involves repositioning of WHD rela-tive to the NBD–HD1 module. Such a switch results inthe dislocation ofWHD and consequent exposure of a bur-ied ADP molecule in the autoinhibited Apaf-1 to solventin the activated Apaf-1 (Fig. 8A). In fact, the nucleotide-binding pocket in the autoinhibited Apaf-1 is sufficiently

spacious to accommodate an ATP or dATP molecule, ex-cept that the binding pocket is blocked by the WHD. Thesolvent-exposed nature of the nucleotide-binding pocketin the activated Apaf-1 may greatly facilitate the displace-ment of ADP by ATP or dATP.

There are at least three prominent differences in nucle-otide binding between the autoinhibited and activatedApaf-1 molecules (Fig. 8B). First, His438 of the MHD/LHDmotif directly coordinates ADP in the autoinhibitedApaf-1 but is distant from the nucleotide-binding pocketin the activated Apaf-1. Second, the conserved Sensor Iresidue Arg265 among the AAA+ proteins is H-bondedto Asp439 in the autoinhibited Apaf-1 but directly coordi-nates dATP by donating a H bond to the γ-phosphategroup (Fig. 8B). Third, the Walker B residue Asp244 isidle in the autoinhibited Apaf-1 but directly contacts theγ-phosphate group of dATP in the activated Apaf-1, per-haps as a consequence of realignment of Arg265. In addi-tion, the phosphate groups of the bound dATP moleculein the activated Apaf-1 are recognized by a number of spe-cific H bonds from Lys160 and other residues from the Ploop (Fig. 8B). Similar to that in the autoinhibited Apaf-1(Riedl et al. 2005; Reubold et al. 2011), the adenine baseof dATP appears to be recognized by specific H bondsfrom main chain groups in the activated Apaf-1.

Discussion

The concept of apoptosis and its role in physiology andhealth were elegantly described in 1972 by Kerr et al.

Figure 7. Conformational differences between theautoinhibited Apaf-1 and the activated Apaf-1 proto-mer from the apoptosome. (A) Structural overlay ofthe autoinhibited Apaf-1 and the activated Apaf-1protomer on their respective NBD–HD1 modules.(B) A close-up view of the two WHDs derived fromA. These two WHDs are related to each other by apseudo-twofold symmetry axis. (C ) Superposition ofthe two WHDs results in near-perfect alignment ofthe WHD–HD2–WD2 rods from the two Apaf-1 mol-ecules. (D) A close-up view of the WD1 and WD2 de-rived from C. Relative to the autoinhibited Apaf-1,WD1 in the activatedApaf-1 promoter undergoes a ro-tation of ∼60° toward the WD2. This movement istriggered by CytC binding. (E) Movement of WD1would cause severe steric clashes with the NBD inthe autoinhibited conformation of Apaf-1. This find-ing explains why, in addition to WD1, CytC bindingmust induce conformational changes in other do-mains of Apaf-1.

Zhou et al.





(1972). Since then, this cellular suicide program has beensubjected to rigorous investigation in severalmodel organ-isms, exemplified by the identification of a linear pathwayof programmed cell death in the nematode Caenorhabdi-tis elegans (Horvitz 2003). The functions of the threeC. elegans genes ced-3, ced-4, and ced-9were clearly delin-eated (Hengartner 1996), with CED-3 being a cell-killingprotease known as caspase (Xue et al. 1996), CED-4 beinga novel adaptor protein required for CED-3 activation(Yuan andHorvitz 1992), andCED-9 being a functional ho-molog of the mammalian oncogene Bcl-2 (Hengartner andHorvitz 1994). The discovery of Apaf-1 as a mammalianhomolog of CED-4 filled a critical missing piece in the de-scription of the conserved apoptosis pathway (Zou et al.1997). More importantly, the identification of the mito-chondrial protein CytC as a cytosolic activator of Apaf-1provides unequivocal evidence for the concept that mito-chondria may play an essential role in the apoptosis ofmammalian cells (Liu et al. 1996; Li et al. 1997). CytCwas subsequently found to work together with dATP/ATP in the assembly of the apoptosome (Li et al. 1997).In the past two decades, numerous advances have been

achieved in the biochemical and structural studies of apo-ptotic mechanisms in several model organisms (Fesik2000; Yan and Shi 2005; Chai and Shi 2014). However, anumber of important questions remain unanswered. The

molecularmechanismofCytC release fromthemitochon-dria is yet to be elucidated. Once in the cytoplasm, howCytC relieves the autoinhibited conformation of Apaf-1and how CytC cooperates with dATP to facilitate apopto-some formation are also largely enigmatic. Addressingthese latter questions requires 3D structural informationof the Apaf-1 apoptosome at an atomic resolution, whichhas remainedelusivedespite repeatedefforts. In this study,we report the cryo-EM structure of an intact Apaf-1 apop-tosome at 3.8Å resolution,which reveals the precise innerworkings of a mammalian apoptosome.Our cryo-EM structure, in conjunction with biochemi-

cal evidence, unveils the underlying molecular mecha-nism for the activation of autoinhibited Apaf-1 and theassembly of a functional apoptosome (Fig. 9). Prior toCytC binding, Apaf-1 exists in an autoinhibited, mono-meric state,which is characterized by specific interactionsbetweenWD1 and theNBD–HD2domains (SupplementalFig. S7). In particular, the previously unnoted cation:π in-teractions between Lys637 of WD1 and Trp249 of NBDare likely to play a major role in locking the autoinhibitedconformationofApaf-1 (Supplemental Fig. S7). In the auto-inhibited Apaf-1, the bound ADP molecule is deeply bur-ied and inaccessible to solvent (Fig. 8), which presumablyblocks potential nucleotide exchange. CytC binds toApaf-1 with a dissociation constant of ∼0.49 µM± 0.06

Figure 8. Comparison of nucleotide binding between the autoinhibited Apaf-1 and the activated Apaf-1 protomer from the apoptosome.(A) An overall comparison of the domains that contribute to nucleotide binding. (Left panel) In the activated Apaf-1 protomer, dATP isbound exclusively by NBD and HD1 and exposed to solvent. (Right panel) In the autoinhibited Apaf-1, ADP is buried and bound at theinterface of NBD, HD1, andWHD. (B) A close-up comparison of nucleotide binding in these two Apaf-1 molecules. (Left panel) In the ac-tivated Apaf-1 protomer, dATP is recognized through a number of H bondsmediated by the Sensor I Arg265, theWalker B residue Asp244,the P-loop residue Lys160, and other main chain groups. No residue from WHD is involved in coordinating dATP. (Right panel) In theautoinhibited Apaf-1, ADP is recognized in part by the MHD/LHD residue His438 through a H bond.






µM, triggering a 60° rotation of WD1 toward WD2. Theconformational switch of WD1 serves two distinct roles:disengaging its interactions with the NBD–HD2 domainsand sterically clashing with NBD of the autoinhibitedApaf-1. Thus, CytC binding primes Apaf-1 for nucleotideexchange (Fig. 9). The replacement of ADP by dATP orATP may stabilize the activated conformation of Apaf-1,thus allowing seven Apaf-1 molecules to assemble intoan apoptosome. Compared with the autoinhibited Apaf-1, the WHD–HD2–WD2 rod in the activated Apaf-1 is ro-tated nearly 180° relative to the NBD–HD1 module.

Our proposed mechanism of Apaf-1 activation andapoptosome formation is consistent with all availablestructural and biochemical evidence. In this mechanism,the primary role of CytC binding is to destabilize the auto-inhibited conformation of Apaf-1 by weakening or dis-rupting the interactions between WD1 and the NBD–

HD2 domains. Truncation of the WD1–WD2 domainsfrom Apaf-1 eliminates these inhibitory interactions,thus allowing the truncated Apaf-1 to form a miniapopto-some that is fully capable of activating caspase-9 (Hu et al.1998; Srinivasula et al. 1998; Riedl et al. 2005). Thismech-anism also predicts inefficient or little nucleotide ex-change in the absence of CytC binding because thebound ADPmolecule is inaccessible to solvent. Althoughsomewhat similar to that proposed by Akey and col-leagues (Yuan et al. 2013), our proposed mechanism isbased on the 3.8 Å structure of the Apaf-1 apoptosomeand structure-guided biochemical analyses and thus con-

tains structurally validated information. For example,the way CytC binds Apaf-1 is different from that proposedpreviously (Yuan et al. 2013), and the observed interac-tions are corroborated by biochemical analysis (Figs. 5, 6).

Dark in fruit flies and CED-4 inC. elegans are the func-tional homologs of Apaf-1 inmammals.We previously de-termined the X-ray structure of the CED-4 apoptosome(Qi et al. 2010) and the cryo-EM structure of the Darkapoptosome (Pang et al. 2015). The advent of the Apaf-1apoptosome allows comparison of the activated confor-mations among Dark, CED-4, and Apaf-1 (SupplementalFig. S8). The Dark protomer can be superimposed toApaf-1 with a root mean squared deviation (RMSD) of∼5.95 Å over 260 aligned Cα atoms (Supplemental Fig.S8A). Despite the moderate RMSD value, the four do-mains NBD, HD1, WHD, and HD2 are reasonably wellaligned. However, the WD1 and WD2 domains in Darkare placed much closer to each other than those in Apaf-1 (Supplemental Fig. S8B); this observation is consistentwith the notion that CytC is dispensable for assembly ofthe Dark apoptosome. The CED-4 protomer can bealigned to Apaf-1 with an RMSD value of 2.71 Å over252 Cα atoms (Supplemental Fig. S8C).

Our structure of the Apaf-1 apoptosome represents asignificant step in an age-old quest to understand how cas-pase-9 is activated. The observed atomic interactions andstructural features in the apoptosome not only explain theassembly of apoptosome but also serve as a molecularframework for understanding caspase-9 activation. The

Figure 9. Mechanism of Apaf-1 activation andapoptosome assembly. In the absence of apoptoticstimuli, Apaf-1 exists in cells as an ADP-bound,autoinhibitedmonomer. At the onset of apoptosis,CytC is released into the cytoplasm,where it bindsto theWD2 andWD1domains of Apaf-1with a dis-sociation constant of ∼0.49 µM. CytC bindingbrings WD1 closer to WD2 and pushes the NBD–

HD1 module away, resulting in the departure ofWHD from the nucleotide-binding interface andsubsequent exposure of the bound ADP to solvent.Such changes may greatly facilitate displacementof ADP by dATP or ATP, which favors the activat-ed conformation of Apaf-1. Seven molecules of ac-tivated Apaf-1 assemble into a closed apoptosome.

Zhou et al.





CARDs, which are essential for caspase-9 recruitment andactivation, are flexible and disordered in the apoptosome.We speculate, as suggested by Akey and colleagues (Yuanet al. 2011), that these CARDs may become ordered andorganized upon caspase-9 binding. These yet to be ob-served features may provide mechanistic insights intoapoptosome-mediated activation of caspase-9.

Materials and methods

Apaf-1 purification and apoptosome assembly

The full-length human Apaf-1 cDNA with a C-terminal 10xHistag was cloned in the pFastBac vector, and the baculovirus wasgenerated using the Bac-Bac system (Invitrogen). The recombi-nant Apaf-1 protein was overexpressed in Hi-5 insect cells (Invi-trogen). Forty-eight hours after viral infection, the cells wereharvested by centrifugation and homogenized in 20 mM HEPES(pH 7.5), 10 mM KCl, 1.5 mM MgCl2, 1 mM EDTA, and 1 mMDTT. Apaf-1 was purified by nickel affinity chromatography(Ni-NTA; Qiagen) and gel filtration (Superose 6, 10/30; GEHealthcare). The apoptosome was assembled by incubating thefull-length Apaf-1 with an excess amount of horse CytC (Sigma)at molar ratio of ∼1:2 and 1 mM dATP.

CytC purification

The untagged, full-length horse CytC was cloned into the pET-Duet vector and coexpressed with the yeast heme lyase CYC3in Escherichia coli BL21 (DE3) cells. All CytCmutants were gen-erated by PCR-based method. The wild-type and mutant CytCproteins were individually purified as previously described (Yuet al. 2001) and used for biochemical analyses.

Caspase-9 activity assay

The impact of CytCmutations onApaf-1 assemblywasmeasuredby the activity of apoptosome-activated caspase-9 activity. Cas-pase-3 (C163A) was used as the substrate. The full-length Apaf-1 at 1 μM was incubated with 0.5 μM CytC, 4 µM dATP, 0.05μM caspase-9, and 14 μM substrate for 60 min at 37°C. The reac-tion was stopped by adding an equal volume of 2× SDS loadingbuffer. The cleavage activities were examined by SDS-PAGEand Coomassie staining.Fluorescent peptide Ac-LEHD-AFCwas also used for the detec-

tion of caspase-9 protease activity. Caspase-9 at 200 nM, full-length Apaf-1 at 400 nM, and 1 mM dATP were incubatedwith 400 nM wild-type CytC or mutants for 10 min at 22°C.Next, the fluorescent substrate was added to a final concentra-tion of 200 μM. The fluorescence intensity was monitored in afluorescence spectrophotometer (F-4600, Hitachi) with an ex-citation wavelength of 400 nm and an emission wavelength of505 nm.

Gel filtration analysis of interactions between Apaf-1and CytC

The full-length Apaf-1 was individually incubated with differentCytCmutants at a molar ratio of 1:2 before loading onto a Super-dex-200 column (increase 5/150; GE Healthcare). The columnwas pre-equilibrated with 100 mM KCl, 20 mM HEPES (pH7.5), and 5 mM DTT.

ITC

ITCwas used tomeasure the binding affinity between Apaf-1 andCytC. All proteins were prepared in a buffer containing 100 mMKCl and 20 mM HEPES (pH 7.5). The concentrations of Apaf-1and CytC were 4 and 60 μM, respectively. The titration was per-formed at 22°C using a VP-ITC microcalorimeter (MicroCal).Data were fitted using the software Origin 7.0 (MicroCal).

EM

Three-microliter aliquots of assembled Apaf-1 apoptosome, at aconcentration of ∼5 μM,were applied to glow-discharged Quanti-foil 400-mesh CuR1.2/1.3 grids. Grids were blotted in a VitrobotIV (FEI Company) for ∼2–3 sec at 4°C with 100% humidity andthen plunge-frozen in liquid ethane. Cryo-EM images of Apaf-1apoptosome were recorded manually on a K2 Summit detector(Gatan Company) in superresolution mode on an FEI Titan Kriosmicroscope operating at 300 kV. A pixel size of 1.32 Å, defocusvalues between 1.4 and 3.0 μm, a dose rate of approximatelyfive electrons per square angstrom per second, and an exposuretime of 8 sec were used on the K2 Summit detector. UCSFImage4was used for data collection (developed by Xueming Li).

Image processing

The 32 movie frames of each micrograph were aligned by whole-imagemotion correction to correct beam-inducedmovements (Liet al. 2013). Contrast transfer function parameters of the resultingmicrographs were estimated by CTFFIND4 (Mindell and Grigor-ieff 2003). Using Relion (version 1.4-alpha) (Scheres 2012),202,932 particles were autopicked from 912 micrographs. Afterthree rounds of reference-free 2D class averaging and one roundof 3D classification, 134,919 good particles were selected for 3Drefinement. A 40 Å low-pass filtered cryo-EM reconstruction ofApaf-1 apoptosome (The Electron Microscopy Data Bank[EMDB] 5186) (Yuan et al. 2013) was used as an initial model inthe 3D refinement. The reconstruction map exhibited an overallresolution of 3.8 Å, with higher resolutions for the central regionbut relatively poor density for CytC and the two WD40 repeatpropellers (WD1 and WD2).By applying a local mask around one spoke and performing 3D

classification without any alignment, particles were classified byonly considering the differences in the spoke region, other thanthe averaged differences from the entire apoptosome. The result-ing class with the largest number of particles was chosen for 3Drefinement. Only this spoke region and the linked central hubwere aligned and refined locally, within 5°. This classificationstrategy proved to be useful for improving the local map quality.Because an apoptosome has a sevenfold symmetry, we rotatedeach particle around the symmetry axis seven times to put all sev-en spokes in the same position for local classification strategy.The original 134,919 particles that had been used to generatethe 3.8 Å density map were treated as the first copy. These parti-cles were rotated 360°/7 around the symmetry axis by adding360/7 to the column “_rlnAngleRot” in the Relion input starfile, resulting in the second copy of the particles. The secondcopy of the particles was similarly rotated by adding another360/7 to the column “_rlnAngleRot,” generating the third copyof the particles. This operation was repeated four more times togenerate the fourth, fifth, sixth, and seventh copies of the parti-cles. For 134,919 particles, 944,433 spokes were rotated to thesame position and classified with the local mask. The resultinglargest class contained 196,815 spokes in the same configuration.The final reconstruction of these particles markedly improvedthe resolution in the spoke region to ∼5 Å, into which atomic






coordinates from crystals structures were docked. Reported reso-lutions are based on the gold-standard FSC 0.143 criterion (Chenet al. 2013). The resulting maps from refinement were post-pro-cessed by Relion for correction of the modulation transfer func-tion (MTF) of the detector and sharpened by a negative B factor(Rosenthal and Henderson 2003). Local resolution variationswere estimated using ResMap (Kucukelbir et al. 2014).

Model building and refinement

The atomic coordinates of Apaf-1 (Protein Data Bank [PDB] code3J2T) served as the initial model. The structure was docked intothe overall map by COOT (Emsley and Cowtan 2004) and fittedinto density by Chimera (Pettersen et al. 2004). Initial structurerefinement of Apaf-1 was carried out by Phenix in real space (Ad-ams et al. 2002) with secondary structure and geometry restraintsto prevent overfitting. During refinement, sevenfold noncrystal-lographic symmetry (NCS) restraint was applied. One Apaf-1 pro-tomerwasmanually adjusted and built usingCOOT, and the finalmodel was generated with a C7 symmetry operator using theCCP4 program (Collaborative Computational Project 1994).The Equus caballus CytC (PDB code 4RSZ) was docked intothe improved density of this region from the local masking map.

Accession numbers

The atomic coordinates of the Apaf-1 apoptosome have beendeposited in the PDB with the accession code 3JBT. The EMmap has been deposited in the EMDB with accession codeEMD-6480.

Acknowledgments

We thank the Tsinghua University Branch of China NationalCenter for Protein Sciences (Beijing) for providing the cryo-EM fa-cility support for image acquisition, and the Tsinghua NationalLaboratory for Information Science and Technology for providingthe “Explorer 100” cluster system for computation. This workwas supported by funds from the Ministry of Science and Tech-nology (2014ZX09507003006 to Y.S.), the National Natural Sci-ence Foundation of China (projects 31430020, 31130002, and31321062 to Y.S.), a European Union Marie Curie Fellowship(to X.-C.B.), and the UK Medical Research Council(MC_UP_A025_1013 to S.H.W.S.).

References

Acehan D, Jiang X, Morgan DG, Heuser JE, Wang X, Akey CW.2002. Three-dimensional structure of the apoptosome: impli-cations for assembly, procaspase-9 binding and activation.Mol Cell 9: 423–432.

Adams P, Grosse-Kunstleve R, Huang L-W, Ioerger T, McCoy A,Moriarty N, Read R, Sacchettini J, Sauter N, Terwilliger T.2002. Phenix: building new software for automated crystallo-graphic structure determination. Acta Crystallogr D58:1948–1954.

Bao Q, LuW, Rabinowitz JD, Shi Y. 2007. Calcium blocks forma-tion of apoptosome by preventing nucleotide exchange inApaf-1. Mol Cell 25: 181–192.

Chai J, Shi Y. 2014. Apoptosome and inflammasome: conservedmachineries for caspase activation. Natl Sci Rev 1: 101–118.

Chen S, McMullan G, Faruqi AR, Murshudov GN, Short JM,Scheres SH, Henderson R. 2013. High-resolution noise substi-

tution to measure overfitting and validate resolution in 3Dstructure determination by single particle electron cryomicro-scopy. Ultramicroscopy 135: 24–35.

Collaborative Computational Project N. 1994. The CCP4 suite:programs for protein crystallography. Acta Crystallogr D50:760–763.

Danial NN, Korsmeyer SJ. 2004. Cell death: critical controlpoints. Cell 116: 205–219.

Diemand AV, Lupas AN. 2006. Modeling AAA+ ring complexesfrom monomeric structures. J Struct Biol 156: 230–243.

Emsley P, Cowtan K. 2004. COOT: model-building tools for mo-lecular graphics. Acta Crystallogr D Biol Crystallogr 60:2126–2132.

Erzberger JP, Berger JM. 2006. Evolutionary relationships andstructural mechanisms of AAA+ proteins. Annu Rev BiophysBiomol Struct 35: 93–114.

Fesik SW. 2000. Insights into programmed cell death throughstructural biology. Cell 103: 273–282.

Hengartner MO. 1996. Programmed cell death in invertebrates.Curr Opin Genet Dev 6: 34–38.

Hengartner MO, Horvitz HR. 1994. C. elegans cell survival geneced-9 encodes a functional homolog of the mammalian proto-oncogene bcl-2. Cell 76: 665–676.

Horvitz HR. 2003. Worms, life, and death (Nobel lecture).Chem-biochem 4: 697–711.

Hu Y, Ding L, Spencer DM, Nunez G. 1998. WD-40 repeat regionregulates Apaf-1 self-association and procaspase-9 activation.J Biol Chem 273: 33489–33494.

Kerr JFF,WylieAH,CurrieAR. 1972. Apoptosis: a basic biologicalphenomenon with wide-ranging implications in tissue kinet-ics. Br J Cancer 26: 239–257.

Kim HE, Du F, Fang M, Wang X. 2005. Formation of apoptosomeis initiated by cytochrome c-induced dATP hydrolysis andsubsequent nucleotide exchange on Apaf-1. Proc Natl AcadSci 102: 17545–17550.

Kucukelbir A, Sigworth FJ, Tagare HD. 2014. Quantifying the lo-cal resolution of cryo-EM density maps. Nat Methods 11:63–65.

Li P, Nijhawan D, Budihardjo I, Srinivasula SM, Ahmad M,Alnemri ES, Wang X. 1997. Cytochrome c and dATP-depen-dent formation of Apaf-1/Caspase-9 complex initiates an apo-ptotic protease cascade. Cell 91: 479–489.

Li X, Mooney P, Zheng S, Booth CR, Braunfeld MB, Gubbens S,Agard DA, Cheng Y. 2013. Electron counting and beam-in-duced motion correction enable near-atomic-resolution sin-gle-particle cryo-EM. Nat Methods 10: 584–590.

Liu X, KimCN, Yang J, Jemmerson R,Wang X. 1996. Induction ofapoptosis program in cell-free extracts: requirement for dATPand cytochrome c. Cell 86: 147–157.

Mindell JA, Grigorieff N. 2003. Accurate determination of localdefocus and specimen tilt in electron microscopy. J StructBiol 142: 334–347.

Pang Y, Bai XC, Yan C, HaoQ, Chen Z,Wang JW, Scheres SH, ShiY. 2015. Structure of the apoptosome: mechanistic insightsinto activation of an initiator caspase fromDrosophila.GenesDev 29: 277–287.

Pettersen EF, Goddard TD, Huang CC, Couch GS, GreenblattDM, Meng EC, Ferrin TE. 2004. UCSF Chimera—a visualiza-tion system for exploratory research and analysis. J ComputChem 25: 1605–1612.

Qi S, Pang Y, Hu Q, Liu Q, Li H, Zhou Y, He T, Liang Q, Liu Y,Yuan X, et al. 2010. Crystal structure of the Caenorhabditiselegans apoptosome reveals an octameric assembly of CED-4. Cell 141: 446–457.

Zhou et al.





ReuboldTF,WohlgemuthS,EschenburgS.2011.Crystal structureof full-lengthApaf-1: how the death signal is relayed in themi-tochondrial pathway of apoptosis. Structure 19: 1074–1083.

Riedl SJ, LiW,ChaoY, Schwarzenbacher R, Shi Y. 2005. Structureof the apoptotic protease activating factor 1 bound to ADP.Nature 434: 926–933.

Rodriguez J, Lazebnik Y. 1999. Caspase-9 and Apaf-1 form an ac-tive holoenzyme. Genes Dev 13: 3179–3184.

Rosenthal PB, Henderson R. 2003. Optimal determination of par-ticle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy. J Mol Biol 333: 721–745.

Saleh A, Srinivasula SM, Acharya S, Fishel R, Alnemri ES. 1999.Cytochrome c and dATP-mediated oligomerization of Apaf-1is a prerequisite for procaspase-9 activation. J Biol Chem 274:17941–17945.

Scheres SH. 2012. RELION: implementation of a Bayesian ap-proach to cryo-EM structure determination. J Struct Biol180: 519–530.

Shi Y. 2002. Mechanisms of caspase inhibition and activationduring apoptosis. Mol Cell 9: 459–470.

Srinivasula SM, Ahmad M, Fernandes-Alnemri T, Alnemri ES.1998. Autoactivation of procaspase-9 byApaf-1-mediated olig-omerization. Mol Cell 1: 949–957.

XueD, ShahamS,HorvitzHR. 1996. TheCaenorhabditis eleganscell-death protein CED-3 is a cysteine protease with substratespecificities similar to those of the human CPP32 protease.Genes Dev 10: 1073–1083.

Yan N, Shi Y. 2005. Mechanisms of apoptosis through structuralbiology. Annu Rev Cell Dev Biol 21: 35–56.

Yu T, Wang X, Purring-Koch C, Wei Y, McLendon GL. 2001. Amutational epitope for cytochrome c binding to the apoptosisprotease activation factor-1 (Apaf-1). J Biol Chem 276:13034–13038.

Yu X, Acehan D, Menetret JF, Booth CR, Ludtke SJ, Riedl SJ, ShiY, Wang X, Akey CW. 2005. A structure of the human apopto-some at 12.8 Å resolution provides insights into this cell deathplatform. Structure 13: 1725–1735.

Yuan J, Horvitz HR. 1992. TheCaenorhabditis elegans cell deathgene ced-4 encodes a novel protein and is expressed during theperiod of extensive programmed cell death.Development 116:309–320.

Yuan S, Yu X, TopfM, Ludtke SJ, Wang X, Akey CW. 2010. Struc-ture of an apoptosome-procaspase-9 CARD complex. Struc-ture 18: 571–583.

Yuan S, Yu X, Asara JM, Heuser JE, Ludtke SJ, Akey CW. 2011.The holo-apoptosome: activation of procaspase-9 and interac-tions with caspase-3. Structure 19: 1084–1096.

Yuan S, Topf M, Reubold TF, Eschenburg S, Akey CW. 2013.Changes in Apaf-1 conformation that drive apoptosome as-sembly. Biochemistry 52: 2319–2327.

ZouH, HenzelWJ, Liu X, Lutschg A,Wang X. 1997. Apaf-1, a Hu-man Protein Homologous toC. elegansCED-4, participates incytochrome c-dependent activation of caspase-3. Cell 90:405–413.

ZouH, Li Y, Liu X,WangX. 1999. AnAPAF-1–cytochrome cmul-timeric complex is a functional apoptosome that activatesprocaspase-9. J Biol Chem 274: 11549–11556.






10.1101/gad.272278.115Access the most recent version at doi: originally published online November 5, 201529:2015, Genes Dev.

Mengying Zhou, Yini Li, Qi Hu, et al. and dATP-mediated activation of Apaf-1

-cAtomic structure of the apoptosome: mechanism of cytochrome

Material

Supplemental

http://genesdev.cshlp.org/content/suppl/2015/11/05/gad.272278.115.DC1

References

http://genesdev.cshlp.org/content/29/22/2349.full.html#ref-list-1

This article cites 43 articles, 9 of which can be accessed free at:

License

Commons Creative

.http://creativecommons.org/licenses/by/4.0/License (Attribution 4.0 International), as described at

, is available under a Creative CommonsGenes & DevelopmentThis article, published in

ServiceEmail Alerting

click here.right corner of the article or

Receive free email alerts when new articles cite this article - sign up in the box at the top

© 2015 Zhou et al.; Published by Cold Spring Harbor Laboratory Press


http://genesdev.cshlp.org/lookup/doi/10.1101/gad.272278.115

http://genesdev.cshlp.org/content/suppl/2015/11/05/gad.272278.115.DC1

http://genesdev.cshlp.org/content/29/22/2349.full.html#ref-list-1


http://genesdev.cshlp.org/cgi/alerts/ctalert?alertType=citedby&addAlert=cited_by&saveAlert=no&cited_by_criteria_resid=protocols;10.1101/gad.272278.115&return_type=article&return_url=http://genesdev.cshlp.org/content/10.1101/gad.272278.115.full.pdf



atomic structure of the apoptosome: mechanism of cytochrome c...

Documents